public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
From: Jiu Fu Guo <guojiufu@gcc.gnu.org>
To: gcc-cvs@gcc.gnu.org
Subject: [gcc(refs/vendors/ibm/heads/perf)] sccvn: Fix handling of POINTER_PLUS_EXPR in memset offset [PR93582]
Date: Thu, 19 Mar 2020 06:07:54 +0000 (GMT)	[thread overview]
Message-ID: <20200319060754.7B1D4394740B@sourceware.org> (raw)

https://gcc.gnu.org/g:fe19699ae2883b252d30f98481d32dabff00744b

commit fe19699ae2883b252d30f98481d32dabff00744b
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Thu Mar 5 08:00:04 2020 +0100

    sccvn: Fix handling of POINTER_PLUS_EXPR in memset offset [PR93582]
    
    > > where POINTER_PLUS_EXPR last operand has sizetype type, thus unsigned,
    > > and in the testcase gimple_assign_rhs2 (def) is thus 0xf000000000000001ULL
    > > which multiplied by 8 doesn't fit into signed HWI.  If it would be treated
    > > as signed offset instead, it would fit (-0xfffffffffffffffLL, multiplied
    > > by 8 is -0x7ffffffffffffff8LL).  Unfortunately with the poly_int obfuscation
    > > I'm not sure how to convert it from unsigned to signed poly_int.
    >
    > mem_ref_offset provides a boiler-plate for this:
    >
    > poly_offset_int::from (wi::to_poly_wide (TREE_OPERAND (t, 1)), SIGNED);
    
    Thanks, that seems to work.
    The test now works on both big-endian and little-endian.
    
    2020-03-05  Richard Biener  <rguenther@suse.de>
                Jakub Jelinek  <jakub@redhat.com>
    
            PR tree-optimization/93582
            * tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
            last operand as signed when looking for memset offset.  Formatting
            fix.
    
            * gcc.dg/tree-ssa/pr93582-11.c: New test.
    
    Co-authored-by: Richard Biener <rguenther@suse.de>

Diff:
---
 gcc/ChangeLog                              |  8 ++++++++
 gcc/testsuite/ChangeLog                    |  5 +++++
 gcc/testsuite/gcc.dg/tree-ssa/pr93582-11.c | 16 ++++++++++++++++
 gcc/tree-ssa-sccvn.c                       | 12 ++++++++----
 4 files changed, 37 insertions(+), 4 deletions(-)

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 5b2e4a83721..897c5986e8d 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,11 @@
+2020-03-05  Richard Biener  <rguenther@suse.de>
+	    Jakub Jelinek  <jakub@redhat.com>
+
+	PR tree-optimization/93582
+	* tree-ssa-sccvn.c (vn_reference_lookup_3): Treat POINTER_PLUS_EXPR
+	last operand as signed when looking for memset offset.  Formatting
+	fix.
+
 2020-03-04  Andrew Pinski  <apinski@marvell.com>
 
 	PR bootstrap/93962
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 95c8710800b..d4e7a3ac690 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2020-03-05  Jakub Jelinek  <jakub@redhat.com>
+
+	PR tree-optimization/93582
+	* gcc.dg/tree-ssa/pr93582-11.c: New test.
+
 2020-03-04  Martin Sebor  <msebor@redhat.com>
 
 	PR c++/90938
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr93582-11.c b/gcc/testsuite/gcc.dg/tree-ssa/pr93582-11.c
new file mode 100644
index 00000000000..9ff400eff1d
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr93582-11.c
@@ -0,0 +1,16 @@
+/* PR tree-optimization/93582 */
+/* { dg-do compile { target lp64 } } */
+/* { dg-options "-O2 -fdump-tree-fre1" } */
+/* { dg-final { scan-tree-dump "return 9223372036854775806;" "fre1" } } */
+
+union U { struct A { unsigned long long a : 1, b : 62, c : 1; } a; unsigned long long i; };
+
+unsigned long long
+foo (char *p)
+{
+  __builtin_memset (p - 0xfffffffffffffffULL, 0, 0xffffffffffffffeULL);
+  __builtin_memset (p + 1, 0, 0xffffffffffffffeULL);
+  union U *q = (union U *) (void *) (p - 4);
+  q->a.b = -1;
+  return q->i;
+}
diff --git a/gcc/tree-ssa-sccvn.c b/gcc/tree-ssa-sccvn.c
index 9ea30ef433c..b7174cd603f 100644
--- a/gcc/tree-ssa-sccvn.c
+++ b/gcc/tree-ssa-sccvn.c
@@ -2656,7 +2656,8 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
 	{
 	  poly_int64 soff;
 	  if (TREE_CODE (base) != MEM_REF
-	      || !(mem_ref_offset (base) << LOG2_BITS_PER_UNIT).to_shwi (&soff))
+	      || !(mem_ref_offset (base)
+		   << LOG2_BITS_PER_UNIT).to_shwi (&soff))
 	    return (void *)-1;
 	  offset += soff;
 	  offset2 = 0;
@@ -2666,10 +2667,13 @@ vn_reference_lookup_3 (ao_ref *ref, tree vuse, void *data_,
 	      if (is_gimple_assign (def)
 		  && gimple_assign_rhs_code (def) == POINTER_PLUS_EXPR
 		  && gimple_assign_rhs1 (def) == TREE_OPERAND (base, 0)
-		  && poly_int_tree_p (gimple_assign_rhs2 (def))
-		  && (wi::to_poly_offset (gimple_assign_rhs2 (def))
-		      << LOG2_BITS_PER_UNIT).to_shwi (&offset2))
+		  && poly_int_tree_p (gimple_assign_rhs2 (def)))
 		{
+		  tree rhs2 = gimple_assign_rhs2 (def);
+		  if (!(poly_offset_int::from (wi::to_poly_wide (rhs2),
+					       SIGNED)
+			<< LOG2_BITS_PER_UNIT).to_shwi (&offset2))
+		    return (void *)-1;
 		  ref2 = gimple_assign_rhs1 (def);
 		  if (TREE_CODE (ref2) == SSA_NAME)
 		    ref2 = SSA_VAL (ref2);


                 reply	other threads:[~2020-03-19  6:07 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200319060754.7B1D4394740B@sourceware.org \
    --to=guojiufu@gcc.gnu.org \
    --cc=gcc-cvs@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).